[plt-scheme] ipc among mzscheme processes

Before I hack up something, is someone already working on a portable IPC
mechanism for MzScheme programs?

One agent per MzScheme interpreter, one MzScheme per host OS process.
Communication across network hosts would be a big bonus.  Must work
under Unix; Mac and Microsoft would be nice too.

I'm strongly leaning towards the old standby of spitting sexps over
TCP/IP sockets (no Unix domain sockets).  Probably one connection per
message.

SSL can be used to encrypt when the agents are on different hosts
Internet, and to authenticate both ends in a portable way.

Should be implementable in a small library that can be easily coded with
in agent programs.  Eventually, the agents will probably be modeled as
concurrent state machines, so the IPC mechanism should allow but not
require this.

Better ideas?

(Rejected: XML-RPC is a ridiculous non-use of XML, and not helpful.
Microsoft lock-in 'standards' are out.  CORBA implementation is too much
hassle for my needs.  Talking HTTP among agents is not a bad idea, but
it's more complexity than necessary with no obvious benefits over
sexp-spitting.)

-- 
                                             http://www.neilvandyke.org/